为解决船舶碰撞事故引发的船体破损、沉没及人员被困问题，提出一种基于改进蚁群算法的船舶机舱救援机器人路径规划方法。针对传统蚁群算法在路径规划中存在搜索效率低、收敛速度慢等问题，对启发函数和信息素更新等要素进行改进。首先，引入自适应迭代权重因子与局部障碍密度修正项，改善蚁群前阶段搜索效率较低的问题；其次，在最佳路径上增加信息素，在迭代后对最差路径进行适当的信息素削弱，遏制劣势路径的影响。然后，引入自适应参数，使得算法在不同迭代阶段对全局最优、迭代最优及最差路径赋予不同权重，从而增强算法的收敛性与鲁棒性。通过上述改进，算法实现了从单纯依靠距离与信息素浓度到综合考虑目标点引力、障碍物排斥、局部环境复杂度及自适应搜索策略的优化转变。实验结果表明，改进算法较传统算法路径缩短10%，冗余转折节点降低65%，迭代次数减少96.4%。实船机舱场景下的验证进一步证明，该方法显著降低了路径规划中的迂回与冗余转折点，加快了收敛速度，可有效避免局部最优陷阱，进而在复杂环境下实现更高效、更稳定的路径搜索。

To address the issues of hull damage, sinking, and personnel entrapment caused by ship collision accidents, a path planning method for ship engine room rescue robots based on an improved ant colony algorithm was proposed. Aiming at the problems of low search efficiency and slow convergence speed of traditional ant colony algorithm in path planning, the heuristic function and pheromone update were improved. Firstly, an adaptive iterative weighting factor coupled with a local obstacle density correction term was introduced to mitigate the low search efficiency prevalent in the algorithm’s initial stages. Secondly, pheromone reinforcement was strategically applied to the optimal path, while appropriate pheromone diminution was executed on suboptimal paths following each iteration, thereby suppressing the detrimental influence of inferior routes. Thirdly, adaptive parameters were introduced to enable the algorithm to assign different weights to the global optimal, iterative optimal, and worst paths at different iteration stages, thereby enhancing the convergence and robustness of the algorithm. Through the above improvements, the algorithm has achieved an optimization transformation from relying solely on distance and pheromone concentration to comprehensively considering target point gravity, obstacle rejection, local environment complexity, and adaptive search strategies. Experimental results show that the improved algorithm shortens the path by 10 % compared with the traditional algorithm, redundant turning nodes are reduced by 65%, and the number of iterations is reduced by 96.4%. The validation in real ship cabin scenarios further demonstrates that the proposed method significantly reduces detours and redundant turning points in path planning, accelerating convergence speed and effectively avoiding local optimum traps, thereby realizing more efficient and stable path search in complex environments.